Implement SkGLContext swapBuffers with fence syncs

Implement SkGLContext swapBuffers with fence syncs

Improves the GPU measuring accuracy of nanobench by using fence syncs.
Fence syncs are very widely supported and available on almost every
platform.

NO_MERGE_BUILDS
BUG=skia:

Committed: https://skia.googlesource.com/skia/+/d416a5b10ff9e6d4f55a1f5b0419722132d68ff3

[Chris Dalton, 2015-06-23 09:01:52]

cdalton@nvidia.com changed reviewers:
+ bsalomon@google.com, joshualitt@google.com

[Chris Dalton, 2015-06-23 09:01:53]

I was unable to get nanobench running on ios, but all other platforms that I
know of have been tested. Are there instructions somewhere to running nanobbench
on ios?

The good news is almost everything supports fence sync in one form or another,
so this should be a pretty solid fix.

[bsalomon, 2015-06-23 14:09:20]

I think if you set your GYP_DEFINES to include skia_os=ios and
skia_arch_type=x86 and then run gyp_skia you can build for the ios simulator.

https://codereview.chromium.org/1194783003/diff/1/bench/nanobench.cpp
File bench/nanobench.cpp (right):

https://codereview.chromium.org/1194783003/diff/1/bench/nanobench.cpp#newcode147
bench/nanobench.cpp:147: bool needsFrameTiming(int* frameLag) const override {
Wondering if this param should be named maxFrameLag?

https://codereview.chromium.org/1194783003/diff/1/include/gpu/gl/SkGLContext.h
File include/gpu/gl/SkGLContext.h (right):

https://codereview.chromium.org/1194783003/diff/1/include/gpu/gl/SkGLContext....
include/gpu/gl/SkGLContext.h:31: bool getGpuFrameLag(int* frameLag) const {
getMaxGpuFrameLag?

I just don't want it to sound like it is returning a current lag value.

[bsalomon, 2015-06-23 14:53:21]

https://codereview.chromium.org/1194783003/diff/1/src/gpu/gl/SkGLContext.cpp
File src/gpu/gl/SkGLContext.cpp (right):

https://codereview.chromium.org/1194783003/diff/1/src/gpu/gl/SkGLContext.cpp#...
src/gpu/gl/SkGLContext.cpp:46: SkASSERT(0 == fFrameFences[0]);
Can we write this class in a way that it will work if fFrameFences were resized?
That is if I changed the header to one fence I'd get 2 frames of lag.

I anticipate that we may find that different values work better on different
configurations.

[Chris Dalton, 2015-06-23 19:34:11]

https://codereview.chromium.org/1194783003/diff/1/bench/nanobench.cpp
File bench/nanobench.cpp (right):

https://codereview.chromium.org/1194783003/diff/1/bench/nanobench.cpp#newcode147
bench/nanobench.cpp:147: bool needsFrameTiming(int* frameLag) const override {
On 2015/06/23 14:09:20, bsalomon wrote:
> Wondering if this param should be named maxFrameLag?

Done.

https://codereview.chromium.org/1194783003/diff/1/include/gpu/gl/SkGLContext.h
File include/gpu/gl/SkGLContext.h (right):

https://codereview.chromium.org/1194783003/diff/1/include/gpu/gl/SkGLContext....
include/gpu/gl/SkGLContext.h:31: bool getGpuFrameLag(int* frameLag) const {
On 2015/06/23 14:09:20, bsalomon wrote:
> getMaxGpuFrameLag?
> 
> I just don't want it to sound like it is returning a current lag value.

Done.

https://codereview.chromium.org/1194783003/diff/1/src/gpu/gl/SkGLContext.cpp
File src/gpu/gl/SkGLContext.cpp (right):

https://codereview.chromium.org/1194783003/diff/1/src/gpu/gl/SkGLContext.cpp#...
src/gpu/gl/SkGLContext.cpp:46: SkASSERT(0 == fFrameFences[0]);
On 2015/06/23 14:53:21, bsalomon wrote:
> Can we write this class in a way that it will work if fFrameFences were
resized?
> That is if I changed the header to one fence I'd get 2 frames of lag.
> 
> I anticipate that we may find that different values work better on different
> configurations. 

Done.

[bsalomon, 2015-06-23 20:04:31]

lgtm, but can you keep an eye on the bots after landing and revert if anything
goes awry?

[Chris Dalton, 2015-06-23 20:17:25]

On 2015/06/23 20:04:31, bsalomon wrote:
> lgtm, but can you keep an eye on the bots after landing and revert if anything
> goes awry?

Sounds good, thanks.

[Chris Dalton, 2015-06-23 20:17:35]

The CQ bit was checked by cdalton@nvidia.com

[Chris Dalton, 2015-06-23 20:17:36]

The patchset sent to the CQ was uploaded after l-g-t-m from bsalomon@google.com
Link to the patchset: https://codereview.chromium.org/1194783003/#ps60001
(title: "windows build")

[commit-bot: I haz the power, 2015-06-23 20:18:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1194783003/60001

[commit-bot: I haz the power, 2015-06-23 20:18:03]

The author cdalton@nvidia.com has not signed Google Contributor License
Agreement. Please visit https://cla.developers.google.com to sign and manage
CLA.

[commit-bot: I haz the power, 2015-06-23 20:23:47]

Committed patchset #4 (id:60001) as
https://skia.googlesource.com/skia/+/d416a5b10ff9e6d4f55a1f5b0419722132d68ff3